DeepMind's Gato AI Model Impresses, But Exaggerated AGI Claims Miss the Point
DeepMind unveils Gato, a multi-tasking AI model that performs over 600 different tasks simultaneously. However, exaggerated claims about imminent human-level AI threaten to overshadow the model's actual technical achievements.
DeepMind unveils Gato, a new AI model that performs over 600 distinct tasks including playing Atari games, captioning images, chatting, and controlling a robot arm. Unlike previous models like AlphaZero that forget previous skills when learning a new one, Gato learns multiple tasks simultaneously and switches between them without losing its previous knowledge.
Despite this genuine technical milestone, the announcement sparks excessive hype within the AI community. A top DeepMind researcher declares that the path to artificial general intelligence is simply a matter of scaling the model up, which leads to breathless media coverage claiming human-level AI is just around the corner.
Experts worry that this kind of feverish discourse does a disservice to the broader AI field by overshadowing other important research areas. While Gato represents a small but significant advance in multi-tasking capabilities, it still performs these varied tasks at a lower proficiency level than specialized single-task models.